Cylindrical and Asymmetrical 3D Convolution Networks for LiDAR-based Perception

doi:10.1109/TPAMI.2021.3098789

	Cylindrical and Asymmetrical 3D Convolution Networks for LiDAR-based Perception
	Zhu, Xinge 1; Zhou, Hui 2; Wang, Tai 1; Hong, Fangzhou 3; Li, Wei 4; Ma, Yuexin5 ; Li, Hongsheng 1; Yang, Ruigang 6; Lin, Dahua 1
	2022-10-01
发表期刊	IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE (IF:20.8[JCR-2023],22.2[5-Year])
ISSN	0162-8828
EISSN	1939-3539
卷号	44 期号:10
发表状态	已发表
DOI	10.1109/TPAMI.2021.3098789
摘要	State-of-the-art methods for driving-scene LiDAR-based perception often project the point clouds to 2D space and then process them via 2D convolution. Although this corporation shows the competitiveness in the point cloud, it inevitably alters and abandons the 3D topology and geometric relations. A natural remedy is to utilize the 3D voxelization and 3D convolution network. However, we found that in the outdoor point cloud, the improvement obtained in this way is quite limited. An important reason is the property of the outdoor point cloud, namely sparsity and varying density. Motivated by this investigation, we propose a new framework for the outdoor LiDAR segmentation, where cylindrical partition and asymmetrical 3D convolution networks are designed to explore the 3D geometric pattern while maintaining these inherent properties. The proposed model acts as a backbone and the learned features from this model can be used for downstream tasks. In this paper, we benchmark our model on three tasks. For semantic segmentation, our method achieves the state-of-the-art in the leaderboard of SemanticKITTI, and significantly outperforms existing methods on nuScenes and A2D2 dataset. Furthermore, the proposed 3D framework also shows strong performance and good generalization on LiDAR panoptic segmentation and LiDAR 3D detection. IEEE
关键词	Convolution Semantics 2-D convolution Geometric patterns Geometric relations Lidar segmentations Semantic segmentation State of the art State-of-the-art methods Voxelization
URL	查看原文
收录类别	SCI ; EI ; SCIE
语种	英语
资助项目	GRF through the Research Grants Council of Hong Kong["14208417","14207319","14203518","ITS/431/18FX"] ; CUHK[TS1712093] ; Shanghai Committee of Science and Technology, China[20DZ1100800]
WOS研究方向	Computer Science ; Engineering
WOS类目	Computer Science, Artificial Intelligence ; Engineering, Electrical & Electronic
WOS记录号	WOS:000853875300070
出版者	IEEE Computer Society
EI入藏号	20213210745098
EI主题词	Optical radar
EI分类号	716.1 Information Theory and Signal Processing ; 716.2 Radar Systems and Equipment
原始文献类型	Article in Press
来源库	IEEE
引用统计	正在获取...
文献类型	期刊论文
条目标识符	https://kms.shanghaitech.edu.cn/handle/2MSLDSTB/135701
专题	信息科学与技术学院_PI研究组_马月昕
通讯作者	Ma, Yuexin
作者单位	1.Chinese Univ Hong Kong, Hong Kong, Peoples R China 2.SenseTime Res, Hong Kong, Peoples R China 3.Nanyang Technol Univ, Singapore 639798, Singapore 4.Peking Univ, Beijing 100871, Peoples R China 5.ShanghaiTech Univ, Shanghai Engn Res Ctr Intelligent Vis & Imaging, Shanghai 201210, Peoples R China 6.Univ Kentucky, Lexington, KY 40506 USA
通讯作者单位	上海科技大学
推荐引用方式 GB/T 7714	Zhu, Xinge,Zhou, Hui,Wang, Tai,et al. Cylindrical and Asymmetrical 3D Convolution Networks for LiDAR-based Perception[J]. IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE,2022,44(10).
APA	Zhu, Xinge.,Zhou, Hui.,Wang, Tai.,Hong, Fangzhou.,Li, Wei.,...&Lin, Dahua.(2022).Cylindrical and Asymmetrical 3D Convolution Networks for LiDAR-based Perception.IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE,44(10).
MLA	Zhu, Xinge,et al."Cylindrical and Asymmetrical 3D Convolution Networks for LiDAR-based Perception".IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE 44.10(2022).